[non sequitur]

purely geometric

I have known that word since I was in Nth grade and I still am not "totally" sure. Like, does it require eyeballs? It'd be interesting to find some concept in topomathology that a blind person couldn't comprehend. Wishful thinking.

Anything that uses a human comprehensible language to describe shapes has got my thumb up it. (e.g. a thing that can be rasterized by a simple algorithm, turn the lights on.)

my thing is definitely not necessarily a thing, well everything is a thing. It's stimulating and I bet ya this stuff exists in pieces, no doubt. For example Real life topologist have modeling software surely. They probably go crazy with the donuts and the coffee cups. And that's the one percent you and I understand. I probably don't even get a percent. I'm just like "cool".

I'm sayin, the Topologist (geometric, algebraic, analytical, computational, reverse cowgirl) ...they have endless results that I betcha are not applied because there's a mishmash of ways of describing 3d things and you have to feed your thing into the thing. If you want to "prove" something you first have to comprehend the picture and we are super confused about that.

Instead, you have lambdas (...) that are executed in-order. They just produce triangles. You can just spray triangles into the universe if you want. Yes they have names like "sphere" and attributes like "radius" but that's hopeless human space. You just want to know if you've got convex, concave, self-intersecting, anything-intersecting, convex for every slice.

So the triangle-iterator is the primitive of the system. You can chain them, you can execute them all at once, but at every step all the topology "middleware" deduces useful stuff. 

And we now know for sure about the poincaré conjecture, ergo something. Maybe manufacturing related.

All this and more likely even exists in a matlab library with an ancient timestamp at oxbridge.

I wonder if there's such a concept in gpu-land. Maybe that's how you load up your triangles.

You could definitely feed it into bevy (et al) efficiently (talking "a frame" here).

WRT CSG, all that stuff can be done. keep track of inside-outside drop the red triangles. Minkowski? -- visit every vertex with your shape.

WRT scale and facet size, a helpful lambda will build a "part" with triangles selected by a cad intern, it can be scaled and everyone can think "mm" as they already do.

I think it's meaningful to require that the triangles be "laid in order", if only because it's pleasing to imagine. A lambda makes a "shape" which is a bunch of connected triangles, which are laid in order, because, you declare inside and outside with the very first triangle.

If you want to scale/rot/trans a shape, scale triangle 0 and start over (or use obvious shortcuts).

...only related to scad because you could use it to implement scad.